One of the primary assumptions associated with many wildlife and population trend studies is that target species are correctly identified. This assumption may not always be valid, particularly for species similar in appearance to co-occurring species. We examined size overlap and identification error rates among Cooper's (Accipiter cooperii) and sharp-shinned (A. striatus) hawks specific to a raptor migration count station along the Pacific Coast of North America. Illustrating the difficulty of distinguishing between these 2 species, we found overlap in 7 metrics among species–sex groups and in 2 metrics between species, and a principal components analysis revealed a continuum of discrete clusters for each species–sex combination in morphospace. Among juvenile hawks (n  =  940), we found the greatest misidentification rate for male Cooper's hawks (23% of the 156 males were identified as sharp-shinned), lesser error rates for female Cooper's (8%, n  =  339) and female sharp-shinned (6%, n  =  246), and the lowest misidentification rate for male sharp-shinned hawks (0%, n  =  199). We observed a similar pattern of misidentification among adult hawks (n  =  48). We attempted to use conditional probabilities (identification rates) from calibration data to calculate the true number of adult and juvenile Cooper's hawks and sharp-shinned hawks. Discrepancies between total number of observed accipiters and estimated number using calibration data suggest that daily observer misclassification rates are higher than misclassification rates estimated from calibration data and prevent correction of the raw data. Our results illustrate the importance of testing for and quantifying observer error in species identification in wildlife census and population trend studies particularly when target species may be easily confused with other nontarget species.